Pre-compute Storage Memory Neural Network System
Summary
The USPTO published patent application US20260093966A1 filed by Swarup Bhunia et al. describing a pre-compute storage memory system for neural networks. The system includes a memory array storing weight activation sets and an address decoder that retrieves pre-computed results matching input operands for use in determining neural network layer activations during inference or training.
What changed
The USPTO published a patent application (US20260093966A1) for a neural network system utilizing pre-computed storage memory. The invention comprises a pre-compute storage memory with pre-computed results corresponding to neural network operations, a memory array for weight activation sets, and an address decoder that generates addresses matching input operands to fetch pre-computed results. The system accelerates neural network inference and training by retrieving pre-computed activation results rather than computing them in real-time.
This is a patent application filing rather than a binding regulatory document. Technology companies, semiconductor manufacturers, and AI researchers should monitor this application for its potential impact on neural network hardware design. No compliance actions are required as patent applications do not impose regulatory obligations on third parties. The application number is 19316443, and it was filed September 2, 2025.
Archived snapshot
Apr 2, 2026GovPing captured this document from the original source. If the source has since changed or been removed, this is the text as it existed at that time.
PRE-COMPUTATION-BASED IMPLEMENTATION OF NEURAL NETWORKS
Application US20260093966A1 Kind: A1 Apr 02, 2026
Inventors
Swarup Bhunia, Baibhab Chatterjee, Atri Chatterjee, Habibur Rahaman
Abstract
A system comprising a pre-compute storage memory comprising a plurality of pre-computed results that correspond to a neural network operation during inferencing or training, wherein the pre-compute storage memory is configured to provide the plurality of pre-computed results as output that is used in determining an activation of a next layer in a neural network; a memory array configured to store a plurality of weight activation sets; an address decoder configured to (i) generate an address in the pre-compute storage memory that matches an input operand, wherein the address corresponds to a weight activation set of the plurality of weight activation sets that matches the input operand and (ii) fetch a pre-computed result of the plurality of pre-compute results from the pre-compute storage memory based on the address.
CPC Classifications
G06N 3/0495 G06N 3/0464 G06N 3/044
Filing Date
2025-09-02
Application No.
19316443
Related changes
Get daily alerts for USPTO Patent Applications - AI & Computing (G06N)
Daily digest delivered to your inbox.
Free. Unsubscribe anytime.
Source
About this page
Every important government, regulator, and court update from around the world. One place. Real-time. Free. Our mission
Source document text, dates, docket IDs, and authority are extracted directly from USPTO.
The summary, classification, recommended actions, deadlines, and penalty information are AI-generated from the original text and may contain errors. Always verify against the source document.
Classification
Who this affects
Taxonomy
Browse Categories
Get alerts for this source
We'll email you when USPTO Patent Applications - AI & Computing (G06N) publishes new changes.
Subscribed!
Optional. Filters your digest to exactly the updates that matter to you.